I think you’re probably missing why it’s exhausting. The problem is not writing English, it’s the rate of change. Programming is meditative, it is a thinking process, the code you output is an artifact of your thinking. Agent-based development… there is no thinking, no meditation, you’re delegating the thinking to a machine, you’re just barking what you want at it, incessantly, endlessly.

For businesses it makes sense to abandon programming in favor of delegating to agents that can do more in less time, but for programmers, it is a loss. Either be a programmer and code, or be a delegator and delegate, you aren’t going to make the life of a delegator suck any less by trying to trick yourself into thinking you’re programming.

I think the reverse direction is more important: taking a massive complex problem/codebase and decomposing it to short pseudocode. Then you could edit the pseudocode and compile it back into the system.

That's the way software engineers working on large projects work anyway: you first gather context on the state of the system and read it at a level you can understand. Then you propose a change on the simplified representation, and then holistically update the machine-runnable format ("implementation").

I'd be interested in tools that formalize/automate this process more.

It is useful to know the level of how "hands-off" to specify directions for each task. With some experience one can learn when to specify the high level requirements, and when to stop and think about the problem at hand.

For example, with frontend, I can get an LLM to design half a page 80% to my satisfaction with two paragraphs worth of a prompt. "It should look like so, and have a text box here, and room for a demo there".

With ML training or backend or user-facing code, I might instead spend a paragraph thinking out my design intentions for a single function or even a single line, more for myself than the LLM. A harness generates a plan based on that paragraph, which one can then comment on and review the pseudocode it provided, ensuring it aligns with expectations.

Lastly have the LLM output some sort of documentation and "here's what I did" after each change. Your final step is to handwrite (paraphrasing what it gave) into any docs or commit messages, and ensure your commits are small enough to keep this maintainable. Paraphrasing the LLM, rather than the LLM paraphrasing you, is helpful to ensure the commit messages make sense to you three months from now.

I'm confused, it looks like you've just written a new terse language that now costs money to compile?

I’ve been thinking about something along these lines for some time. I really like the direction of this.

The challenge I see more broadly is we (as engineers now empowered by LLMs) are trying to find the right level of abstraction to operate in. Writing long form sentences and (sometime) reviewing the output feels too far away. But having an LLM work directly with you in an IDE feels too close to “the old way”.

Personally for me the approach here still feels a little too close to the lower level old way, but it’s better than the two approaches above.

Excited to see where you take it!

As I see it, there's a bit of internal contradiction: your declared intent is to not write code, but you had to circle back to code (albeit more relaxed and vague) because human English is (compared to proglangs) imprecise. Pseudocode, however, is not far away from it, still not strict, and LLM is still a stochastic generator. So it will continue to randomly deviate from what you want it to do. I guess it could be an enhancement, but who knows, maybe in a year from now, you will get tired with pseudocode not being precise, and go back to writing code :wink:

Dumb question:

Why not just put an instruction into your favorite harness’ system prompt: “If I give you pseudo code, spell out my intent, and then write and test it in real code.”

As I was reading and seeing the video, which is easier to demonstrate as you guessed, I was asking my self the same questions you answered in the "Caveats" section.

I love that you are trying something and you shared it, I relate very much to some parts of the post minus the fatigue part maybe I am it's because I am a very chatty person in general so prompting is not a problem who knows!

I had my own attempts to improve working with AI but to my fault I rarely commit to a project no matter if I wrote it or AI wrote it for me!

I tried this over a year ago https://github.com/ramigb/promachos (before I found out about spec kit and similar solutions) Then I tried this https://github.com/ramigb/groundcheck recently which is to actually help me in the review process specially if there is intent documentation like ADRs or similar.

Your post inspired me to try from a different angle. Thank you.

I'm surprised it wasn't mentioned yet, but it seems pretty similar in concept to codespeak https://news.ycombinator.com/item?id=47350931

Though catching back up on that project- it seems it's evolved pretty substantially, becoming much higher level than the initial pseudocode driven version I remember

I'm not sold on the pseudocode approach, but I agree with the declarative aspect. Declarative specs have become central to my process and I've built this tool to support it:

https://github.com/spekk-ai/spekk-cli

Rather than writing exhaustive specs, I preserve only the intent and what must be true as discrete assertions. This preserves the leverage you get from LLMs - anything it can reliably infer does not need to be specified. It also (mostly) separates intent from code or architecture decisions, which keeps specs flexible.

I really like the “persisting of intent” part of this approach. I don’t have any experience with coding agents, but it seems to me that this could make it much easier/quicker for others to understand an AI-written codebase (or even PR) and maybe also for agents to better pick up on the “human context” of a project.

Personally, I’d feel much more confident to use a vibe-coded library where I can read the human-written intentions than one where I just see a lot of AI-written code. Maybe it feels more like there is still a programmer/engineer/architect behind it who knows what they’re doing and what exactly they want to achieve.

To me, vibe-coding always feels a bit like this weird transition into something else that is more precise and reliable. It feels backwards to give up on our achievements in formal language for the comfortable vagueness of everyday speech.

Are there examples of how this would work when you need the pseudocode to reference abstract application concepts?

I'm not totally convinced this is a useful way to express something like "Change the data flow so that we bulk query from the DB upfront and pass it down to all callsites"

The downside is that... you don't get to brainstorm with the agent about ways to implement stuff. Architecture, design etc. It would feel like.. you're on your own. Of course you could just a regular chat interface alongside but theb you deal with two interfaces. Maybe this could be a target format for another chat agent that you talk to.

I think the idea of having a human-written persistent document describing the operation of the code is a great idea. This document acts as the prompting interface instead of the chat window and changes can still be tracked. Surely something as simple as a skill.md can be made for such a setup, right? I think the pseudocode style is a seperate axis to this setup.

Isn't this just spec-driven development in a different language?

I like the pseudocode approach I think there is some gold there. I been thinking a lot about it (https://www.generativenetworks.com/blog/do-we-still-need-to-...) but working on small apps I never need it only in bigger apps or when you need an extra level of precision in the prompt

A do-what-I-mean interface using pseudocode seems like an interesting idea to explore, but perhaps it should be automatically reformatted by the AI to conform to some grammar? The idea would be a documentation standard (like Markdown), not an actual programming language.

Sometimes you might also want examples and then BDD testing software (like Yadda) might make sense?

Hello. If I may, the way I solved your problem of "you lost the prompt" and "prompts describe changes not state" is by using a technology which was actually born 15 years ago: Behaviour Driven Development. I write (or make the AI write but then I review them carefully) features files with the Gherkins syntax. Then I ask the AI to implement the tets and then make them green. It works very well because they are both a description of the application and actual executable tests so they also spot regression. It's a very nice way to work with AI.

I think the way I would want to work with AI for web apps would be like this:

You install a fancy chrome extension or custom browser.

You go through the app on this browser. You notice something you want to change. You can then submit a prompt via this extension saying the change you want.

The trick is, the chrome extension has been following your movements through the app. This way the chrome extension can generate a lot of data the AI can use for a good prompt. I.e. The extension can get a screenshot of where you are in the app and all the pages you went through to get there. The extension can get the console logs and traces and stuff.

So with all this, the AI system gets all the material needed for a good prompt to make a good change.

In an ideal world, the AI system could then save all these details and when the change is made, guide you through the UX again. And you can check if the change was made as you wanted.

So kinda like AI flavored manual UX testing where you click around and record your findings.

Hmm. Interesting idea.

What about multi-file / larger changes? How would you express files being connected, imports, and exports? Or are you thinking the hz files are disposable per change?

I like the direction of capturing the human intent as a durable artifact, but I dislike how you've gotten there.

Let's look at your fizzbuzz example. Unfortunately, if you wanted to have the agent implement fizzbuzz for you, it looks like, in your example, you would have to already know how to effectively write fizzbuzz. Specifically, you call out the use of the modulo.

In your prompt, for the traditional agentic development path, you already declared the intent. There is some imperative language in there, sure, "Create a function that ...", but also there is the declarative state, that doesn't require knowledge of specific programming syntax or semantics.

What I've relied on is a more formal location/syntax for acceptance criteria are in code. These are then used to generate tests, and implementations. It isn't perfect, and more investment is needed, but it starts getting at the root of the problem.

I understand your approach and applaud it. I have been doing two things that keep me doing the parts I love. Instead of pseudo code I write in a simple managed language like JavaScript or Python. My instructions to the agent is simple. Code gets no comments. It is intentionally brief and for me at least understandable. Second. I have a document for each code file that holds the an enumerated list of rules used to develop the file - The comments if you will but in a format I can deal with line by line, just like the code. This is substantially more rewarding for me as I change both the code and the document and limit terminal interactions to must-see/do stuff.

This approach reminds me of PDL (Program Design Language), described in Steve McConnell's excellent book Code Complete (1993). He recommended writing code using PDL pseudocode first. People can review your PDL before you write the code implementing the PDL, leaving the PDL as code comments.

https://en.wikipedia.org/wiki/Program_Design_Language

https://codecourse.sourceforge.net/materials/Code-Complete-A...

I think its an interesting approach, much more than most other new agent harnesses at least.

I guess you still need the chat to discuss with the AI? For example, ask it to compare solution A and B, or to explain how to do X. A bit like you can use plan mode today. Then given the result, you can write the hz file (or even have an agent write it).

But I wonder how it works on a more complicated project than fizzbuzz. Can I suggest you use Huzzah to develop itself, and then share the hz file(s)? That will show both how it works on a more complicated project, and something that is developing over time.

I think "the pseudocode is persisted alongside the generated code" just reinvented jira/linear tickets and PR descriptions. We have ways of using git and tracing the code write to the thought process behind it.

What I find interesting about this + a random bucket of associations because I’ve fallen under the spell of Satan’s Lettuce:

Just a few days ago someone was talking about a machine - human patois.

This (your project) sits somewhere between Lean and BDD cucumber syntax.

At the same time Claude spits out phrases like “a container paying the price of -42px”.

Recently I was listening to a lecture about metaphor in poetry, the misconception that poems are riddles whereas we use metaphors all the time in our language because they convey the meaning more precisely.

I think that this would be easy to criticize without interfacing with their underlying idea here.

It's cool that with a tool like this you don't NEED to get all aspects of your code finalized and ready. It's possible to be vague when you want to and specific when you need to.

I'm not sure if that itself would work well in practice, but the project is still quite cool nonetheless.

I agree with the intent (and I am also exhausted), but not sure Huzzah is needed. you can already do this using any language or no language. I can write some quick python pseudocode that demonstrates what I want but probably is broken, and then ask Claude/etc to write real code in X language that actually works and is more elegantly or efficiently written than my example.

I could also use a Cucumber (or your favorite BDD) approach, writing out user stories and then have Claude build the code that satisfies them

But this primarily works for smaller self-contained pieces. If you're working on a larger code base with knock-on effects, you can't express it in pseudocode -- unless you maintain a whole "pseudocode twin" of your codebase, which seems like a lot of extra work.

this is a question whole software industry is trying to solve. "how can we make sense of the ai generated codebase"

what you did is spec-driven development, instead of use-cases and requirements you have pseudo-code.

spec-driven did not work in my case, likely yours suffer the same "issue of walls of text no one wants to read".

This is a step in an interesting direction for many reasons. First of all it reduces the "conversation in english" aspect and goes back to kinda "writing code" which I think would reduce a lot of fatigue and bring back some joy in making software. But then it also gives more control over the output in a way that makes sense: I know how i want to code this, but I can save time not having to deal with the syntax/boilerplate/actual writing. Also since I structured everything precisely, reviewing is going to be way easier. It requires though yet another mental shift in how to code: you stop chatting and you go back to writing in a text editor similarly to what you did before, the difference is that now you write some kind of scaffold instead than the actual code.

It has two problems though: if my approach has flaws the agent would implement it as-is even if could instead suggest an improvement. Also, an advantage of agents in huge codebases is that they can find where to make the change and draft it, which wouldn't work with this system.

When the project scales, the transformation of intent into such pseudocode becomes a big deal. Having the pseudocode as an intermediate level to check agents' artifacts is interesting media, but I couldn't see how this would actually work with Huzzah tbh.

If the pseudocode is precise, what you want is a compiler. Otherwise the LLM is still making decisions for you.

I've settled into a habit of asking codex to write down summary of key decisions in a ledger as I end a unit of work. It keeps iterating natural but maintains a system of record on decisions in the same repo. The "ledger" is the new code.

Writing fizzbuzz requires that you understand algo + you credit card, while agent requires only your credit card. I believe most of people will pick the 2nd one.

Are we supposed to be able to read the examples with our eyes? It looks like black text on a very dark grey background on my iPhone.

EDIT: same on Firefox on my Mac (macOS Ventura).

I like the distinction between imperative and declarative styles, but I still use imperative chat sessions to work through what the declarative plan should look like. I feel like this approach loses that.

curious if you’ve tried Kiro or spec-driven development? that seems like it would solve at least some of the issues you raised with agent based development, albeit in a different way without the emphasis on pseudo code

the endgame is a team whose git history is all generated commits, while the one file that actually captures intent is a terse text nobody thought to commit.

It was a very nice post, at least I learned that I was not the only one in a vacuum in this regard. I will try Huzzah and share my views under this topic.

You could write that pseudocode as a prompt for the agent and get the same result. Use plan mode to understand what agent wants to do.

Am i missing anything?

This is basically a compiler, but we’re moving up a layer of abstraction.

Definitely an approach worth exploring! I actually started to look into semi formal spec language like Quint because I wanted something more structured then prose, so I feel like this goes into the right direction.

I had this idea 25 years ago, except that I wanted to use the AI available at the time, which would have probably been genetic algorithms. The idea was to write the expected outputs for some input examples and have a Lisp-like language use evolution/mutation to generate a paragraph or so of code that meets the spec.

Then probably do some fuzzing or SAT solving to formally enumerate the known-good cases and possible edge case exceptions. That was before I knew about ranged variables, category theory, etc.

Unfortunately the real world got in the way, and I spent a quarter century treading water to survive, which was exacerbated by CPTSD amplifying ADHD and OCD symptoms many fold before I knew what they were. So now any great ideas I might have had are rendered obsolete because an LLM can synthesize them in a matter of minutes. It may be time to let go and hand things over to the next generation.

-

I think maybe the goal now is to keep building on these ideas until we achieve creating digital assistants that can manage all of the aspects of our lives that interfere with our goals. That's a controversial statement because the political right doesn't distinguish between setbacks and goals (it's all discipline) while the political left embraces a kind of learned helplessness as controlled opposition. Change comes glacially or not at all. In other words, help isn't coming from above - we have to pick ourselves up by our bootstraps like we always do.

A few things that AI might help solve by rendering the conditions of their existence obsolete:

- Money

- Hunger

- Illness

- Exploitation

- Pollution

The list goes on. Unless we're actively working to solve these things, then whatever we come up with is frankly a distraction. I consider just about all tech innovation since the 1990s to be a waste of time for that reason. It all just made some guy rich. Yawn.

Huzzah however, is great. I hope it leads to something wonderful.

Interesting idea! I wonder if, after some iteration, a variation on this could help curb some of the LLM spaghetti mess

Very insightful. The idea of using pseudocode is clever, and the persistence+correlation layer idea is great.

not sure if the hz file artifact is needed, you can enter pseudocode directly into chat or even on an existing code file and with minor comment agents will be able to work with it. i write this type of pseudocode to existing code files often to great results.

> There’s no reliable record of human intent.

Every engineer I have ever mentored got a lesson on how to write a good commit message that included this. This is exactly that.

Further Huzzah from skimming it over seems to be re-inventing documenting your code.

Together I can only surmise that the author is new out of school or has simply not yet worked on a team with good coding practices.

Are you using Huzzah to development Huzzah? :)

Terse pseudo code > verbose prose

Coding Encoding Think about the terms

i actually really like this. nice job

You seem to be conflating two things: how to prompt, and how to share sessions. You can already use pseudo-code today if you want to. As for sharing, you can commit (a link to) it, use `git notes` (as I do), or a service like entire.io.

I think you should work on your differentiation. The session management stuff is the greater concern, in my opinion; pseudo code is not a novelty.

How is it different from https://github.com/github/spec-kit?

It takes a great deal of bravery to publish work like this. I see the comments are filled with people who have never used an agent to develop code and are quite sure this is the dumbest thing they've ever seen.

Nice work.

I'd call this one Micropilot.

As in micromanagement.

There's a problem being missed here. You're trying to make a big complex thing from a little simple thing. It's like trying to take a hash and turn it back into the original data, without knowing what the original data looked like.

You can get close, but things will be missed. So you'll edit, or provide more info to the AI outside your pseudocode, and maybe get to what you want. But another change will undo the fix you made, because whatever hidden AI context was needed for that first fix is gone now.

The pseudocode isn't a bad idea in terms of trying to visualize a complex idea. But just like writing code, humans suck at truly groking all the implications of their complex ideas, and make mistakes. We need formal methods to double-check those complex ideas to avoid the mistakes. Some kind of formal representation of your idea, in something like a cyclical graph, could have checks on it, to make editing it less error-prone. But you need to get the human out of the loop, or you'll be trapped in the same world we've always been in.

I’ve not tried it but I’m willing to bet if you gave Claude or ChatGPT similar pseudo code you’d get the same result.

Invented New coding language that transpile to other coding languages and saying it novel approach.

I really feel the pain here, I have myself tried things ol these lines... I will give Huzzah it a try.

My concern, however, is that all of these options might ultimately be slower than just prompting.

I mean, at this point, another option would be to just go back to actually writing the code ourselves, like we did back then in 2024 ??

Can't already just write pseudocode in your prompts without any extra tooling?

I'm rarely vibing at the individual function level these days. If I was I'd probably use this to do the unit tests and then get the LLM to write the function from that.

Also, it seems one could get Claude to do this in 2 mins without making a whole IDE specifically for it?

Now days I use the LLM wiki idea to build a detailed spec up front before getting the agent to build the system. Keeps a record of the intent and you can get the agent to keep this updated with every change request.

Another LLM could use the pseudocode to generate black box test cases without access to the generated code. How to define the API between tests and the implementation would require some more work.

I was already doing this with "writing" tests. Just being absolutely terribly lazy.

Nice work!

I wrote this but as a compiler. It was ~2 years ago and local models have gotten WAY better; I was having too many issues with adherence (syntax errors, etc) and dropped it.

The compiler comes with a model embedded or can use an external model. It uses Cosmopolitan Libc and can zip things together into one binary. I will take some time to dust it off and share it.

But the idea was basically, you have your natural language source files or a one-shot prompt and it "compiles" them into a single, shareable fat binary that works across all popular platforms and architectures.

It was pretty fun to use with remote frontier models but the local model story simply wasn't good enough at the time for me to feel proud releasing it. I think that's probably changed now and passable results can be had even with small modern 7B/14B models.

I just use a lot of folders, like folder for ever class and manager class with it and anything else. So its kind of organized by folders like plugins. Much easier to reason about. The biggest problem is the central class that glues everything is now 30 thousand lines, but its mostly repetitive delegation, init update, draw, delete. Still should be some nicer solution than having a huge class in center.

Cheers!

You are like that bad manager who will hire 10x super engineer and then stifle their work with useless policies and meetings and team building activities “to make sure the productivity are better and work is consistent and team harmony is there!!”. The whole idea of using multi billion weight model is you don’t restrict it with your limited knowledge, you only guide it and review after to make sure it aligns with your goals, definitely the model will bring new tools or tricks you never knew it existed let alone they are useful, just like that super engineer doing things on their own approach, you only guide and align to the goal, here in building your software and in the company to your business goals.

It took you 8 months to burn out from churning SaaS slop? My mind was cooked in 3 weeks trying to six neglected ideas at once

> utterly exhausted

Keep your toolchain as simple as possible.

A really nice rig, which you can use in an existing repo, is to have ollama and aider simply log everything that happens in the session, through tee, into a log directory which you do - indeed - check into the repo.

> almost exclusively with coding agents

Do your own commits too (don't just let the ML do them), and in those commits, keep your prompts.

Learn to use your AI skills with succinct and calculated, forthright projection.

Which is to say, it is your own personal set of words now which define your control over your computer.

The words are tools. But what are your methods?

> .. tedious to write full sentences for every change I want .. interaction paradigm ..

Your own command of your speaking/thinking language can be extended as far and as wide, now, as you can possibly imagine. In fact, you must control AI/ML with imagination now, in multiple ways.

One of those ways is to iterate on expansion of your own ontology. There has to be an input from the AI before an adequate human output can send the AI directly at the heart of it. This improvement loop is on you. Get smarter with the loop.

> pseudo-code -> sync -> record of intent

[1] "Idea -> Description -> Result -> Build -> [human] (use)"

Well, I get this by checking all my aider logs into a submodule of my main source tree. All my prompts, all the happy little mistakes and bright, shiny things, commit by commit. Sure, the logs grow and grow, but you know what .. I learn a hell of a lot by reading them.

Time-stamped. So, nice graphs if I wanted them, one of these days we'll do it, me and the AI.

The commit point for where I cut the exhaustion between me and the immense power of the AI/ML tooling, is when there is a new build, and I have tested it, personally.

I get exhausted if there is no delivery factor, to me personally, from whatever method I'm wrangling the tools with. Like if I really push too hard on the prompt, things get gnarly.

But, I've been here before over the decades, there are methods.

Even in the AI/ML age .. tooling and methodology requires a discipline - what is true now more than ever is that if a method fails, the usual approach of building another tool is not necessarily the best approach.

Methods can be sharpened just like tools. But every tool carries a cognitive load.

The methods are there to make that load useful. Are you a user?

So then just do a build and run it. See if is worth it.

Goto [1].

what is this!